Study guide
Hardware, Metadata, and Config Files
Datadog Fundamentals Study Guide Exam
Track your progress as you read
Finished sections get a check mark.
1 Computer Fundamentals
Datadog Cert Format and Study Map Shell, OS, and Processes Networking Basics for Observability Hardware, Metadata, and Config Files Linux/Windows Troubleshooting Basics
2 Infrastructure Development
2.1 Agent Install and Architecture 2.2 API Keys and Application Keys 2.3 Running the Agent and Status 2.4 Hostnames and Environments
3 Networking and Agent Configuration
Datadog Ports and Endpoints Datadog IPs and Connectivity Autodiscovery Patterns
4 Data Collection
DogStatsD and Custom Metrics Integrations and Crawlers API Endpoints for Data Tagging Best Practices Metrics, Timeseries, and Types
5 Troubleshooting Datadog
Agent Commands and Flare Agent Logs and Debug Config File Troubleshooting
6 Data Visualization and Utilization
6.1 Host Map and Infrastructure List 6.2 Dashboards and Widgets 6.3 Monitors, Alerts, and Notifications 6.4 Investigation Workflows and Exam Tips
Hardware, Metadata, and Config Files
Key Takeaways Explain This Topic AI
CPU, memory, disk, and network metrics describe hardware utilization; sustained saturation on any resource can degrade applications the Agent monitors.
Metadata (tags like env:prod or role:database) describes entities for filtering; timeseries metrics are numeric values that change over time.
The main Linux Agent config file is /etc/datadog-agent/datadog.yaml; integration configs live under conf.d/ subdirectories.
YAML is whitespace-sensitive—tabs and inconsistent indentation commonly prevent the Agent from parsing datadog.yaml.
Least-privilege file permissions on config files protect API keys from unauthorized local users.
Last updated: July 2026
Hardware Metrics: What Host Dashboards Really Show
Before you customize dashboards, Datadog collects host-level hardware metrics through the Agent's system checks. These timeseries reflect physical or virtual resource usage:
- Metric family What it indicates Exam trap to avoid
- CPU utilization Processor busy percentage High CPU with low load might still be fine for batch jobs
- Memory used / available RAM pressure Memory exhaustion can slow apps while CPU looks idle
- Disk usage / in_use Filesystem capacity Full disks break logs and Agent buffers
- Network bytes sent/received Traffic volume High traffic is not automatically an error
- Load average Run-queue depth (Unix) Load is not a percentage—context matters
Sustained saturation on CPU, memory, or disk correlates with incident scenarios: slow databases, Agent check timeouts, or missing logs because disk writes fail. The exam may ask which metric you inspect first for a given symptom—match the metric to the resource bottleneck , not whichever graph is prettiest.
Virtualization adds nuance: CPU steal time on oversubscribed hypervisors can throttle guests even when guest CPU graphs look moderate. You do not need deep hypervisor expertise, but recognize that host metrics contextualize application slowness .
Metadata vs Metrics: Two Different Languages
Timeseries metrics are numeric measurements indexed by time— system.cpu.user , system.mem.used , system.disk.in_use . They support aggregation: averages, maxes, rates.
Metadata describes identity and context without being a plotted value itself. In Datadog, metadata often appears as tags attached to hosts, metrics, logs, or traces:
- env:staging — deployment environment
- role:database — functional role
- team:payments — ownership
- os:linux — operating system family
An exam favorite: which item is metadata? role:database is a tag (metadata) ; system.cpu.user is a metric . Tags filter and group ; metrics measure .
Cardinality caution
Tags should come from bounded sets . Using user_id:984302 or request_id:abc creates unbounded cardinality —too many unique tag values. Datadog may struggle to aggregate, dashboards slow down, and billing for custom metrics can spike. The Fundamentals exam tests whether you recognize good tags (env, service, role) vs bad tags (unique per request) .
- Configuration Files: Layout and Purpose
- On a standard Linux install, the Agent's main configuration file is: /etc/datadog-agent/datadog.yaml
This file holds global settings: api_key , site , hostname , tags , proxy configuration, and feature toggles. Integration-specific configs live under: /etc/datadog-agent/conf.d/<integration>.d/conf.yaml
Windows paths differ ( %ProgramData%\Datadog\... ) but the same conceptual split applies: one main file plus per-integration snippets .
When a stem references "main Agent configuration file on Linux," /etc/datadog-agent/datadog.yaml is the expected answer—not log paths, not random JSON under /opt .
Editing YAML Safely
Agent configuration uses YAML , a human-readable format built on indentation instead of braces. Rules that matter on exams and in production:
- Indentation defines nesting —usually two spaces per level
- Tabs are dangerous —many parsers reject tab-indented YAML
- Comments start with # and are ignored
- Strings with special characters may need quotes
- Duplicate keys —last wins or parse fails depending on parser settings; treat as errors
- Example fragment: api_key: "<YOUR_API_KEY>"
- site: datadoghq.com
tags:
- env:prod
- team:checkout
- logs_enabled: true
If an engineer pastes a tab before tags: , the Agent may fail to start entirely —a config parse error, not a network issue. Always run agent configcheck or agent status after edits (Chapter 5 goes deeper).
JSON in Datadog Workflows
JSON appears heavily in API automation —creating monitors, dashboards, or querying metrics programmatically. Unlike YAML config files, JSON uses explicit { } brackets and "key": "value" pairs. Datadog API responses are JSON documents you parse in Python, curl, or Terraform providers.
Do not confuse the formats: YAML for Agent files on disk , JSON for HTTP API bodies . An exam distractor might show JSON syntax inside datadog.yaml —invalid unless properly embedded as a string.
Environment Variables and File Permissions
Secrets such as the API key can live in datadog.yaml or arrive via DD_API_KEY . Regardless of storage, apply least privilege file permissions—typically root or the dd-agent service account can read datadog.yaml , not world-readable chmod 644 on shared servers with untrusted users.
World-writable config ( chmod 777 ) is never correct—any local user could hijack the key or break settings.
Hardware Metadata from the Cloud
Cloud providers expose metadata endpoints (AWS instance ID, Azure VM name, GCP zone). Datadog can enrich hosts with these attributes as tags. That is still metadata , not a replacement for system.cpu metrics . Questions may ask how to group all database hosts — role:database beats instance-id:i-0abc123 for human dashboards, though instance IDs are fine as supplementary tags.
Worked Scenario
Problem: After adding logs_enabled: true to datadog.yaml , the Agent fails to start.
Investigation path:
- Check YAML indentation under the new key
- Validate there is no tab character introduced by the editor
- Confirm logs section syntax matches documented structure
- Read Agent startup logs for parse errors
Lesson: Config-file modification questions often test syntax and placement , not whether logs are a good idea.
Takeaway Table
- Concept Example Used for
- Hardware metric system.mem.used Alert on memory pressure
- Metadata tag env:prod Scope dashboards
- Main config /etc/datadog-agent/datadog.yaml Global Agent settings
- Integration config conf.d/nginx.d/conf.yaml Single integration tuning
- Secret delivery DD_API_KEY env var Containerized deploys